Data-driven prediction of rail neutral temperature for continuously welded rails using rail vibration resonance frequencies

Thermal buckling of continuously welded rail (CWR) has been a long-standing challenge for the railroad industry because of the high derailment rate and the associated social, economic, and environmental impacts it causes. Rail buckling is generally attributed to excessive thermally-induced axial compressive stress developed in the rail from high temperatures. Knowledge of the rail thermal stress, or its rail neutral temperature (RNT), is critical for safe and efficient rail system operation. There has been great interest and much work on the development of nondestructive evaluation (NDE) techniques to estimate rail thermal stress/load and RNT in situ; however, existing techniques do not satisfy both practical application and established accuracy issues at the same time. In this thesis the use of impulse-driven multi-resonance rail vibration is evaluated for estimation of effective axial stain and RNT. The multiple resonances generated from an impulse event provide rich information from a single test; I hypothesize the correlations between the resonance frequencies and rail axial stress condition can provide estimation of RNT without reference measurement or models. This approach also aims to provide estimations while minimizing influences from the environment and conditions of rail supporting structures, which are known to significantly influence measurement of existing techniques. Preliminary results from laboratory tests demonstrate that impulse-generated vibration resonances show measurable and varying levels of sensitivity to, individually, stress and temperature changes in rail. Field measurements collected from two separate test locations on revenue in-service CWR during nine separate testing days over a period of nearly two years show that rail vibration resonances vary with rail temperature and rail stress conditions. This study uses a data-driven approach to investigate correlations between selected resonance frequencies and the in-place effective axial strain or RNT conditions and rail temperature. All excited resonance peaks are first distinguished from other peaks caused by noise using spectral amplitude variance. Among these resonance peaks, potentially useful resonances are identified by evaluating with respect to stacked spectra collected across a testing day using an assumed frequency-temperature relation. A subset of the identified useful resonances is then identified based their consistent appearances across both testing locations and all testing days, strong correlation to effective strain, and strong correlation to each other. Three resonant modes emerge from this process as best candidates. A classic feature selection technique, Lasso linear regression, is then employed to identify critical power combinations of the three resonant mode frequencies. Several of the power combinations exhibit unique correlation to the measured equivalent axial strain at both test locations across all testing days. Two power combinations show ability to predict RNT. The predicted RNT is predicted at one test location using different models based on the power combination data from the other test location, and vice versa, where the predictions satisfy the standard RNT measurement accuracy expectations.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Civil Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2023

Author dc:creator

• Huang, Chi-Luen

Contributors dc:contributor

• Popovics, John S
• Weaver, Richard L
• Meidani, Hadi
• Edwards, J. Riley

• Axial Strain
• Nde
• Ndt
• Cwr
• Lasso Regression
• Machine Learning
• Rnt
• Thermal Buckling

Statement dc:rights

• Copyright 2023 Chi-Luen Huang

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/122261

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F122261